Modelling Wetland Eco-hydrological state in Pressure-State-Response framework validating it using Ecosystem service potentiality approach


 The present study has tried to model the Eco-hydrological state of wetlands for both pre dam and post dam periods of the Lower Tangon river basin using Pressure-State-Response (PSR) framework with advanced machine learning (ML) algorithms and validated it with Ecosystem service potentiality (ESP) approach along with conventional approaches of validation. All the applied models have explored that 22.48–39.52% wetland under very good EHS zone has converted into 15.52–16.57% relatively lower category of EHS zones indicating the gradual degradation of EHS quality over wider parts of the wetland. The result of model validation has noted the acceptability of all the applied models but performance is found high in the case of REPtree and Bagging models. Expert-based ESP behaves accordingly with the EHS models. Based on the results, the study suggests using ML models for such modelling and used field-based validation approach like ESP.


25
Wetlands is a valuable natural resource that covers 6% of the world's geographical surface   160 Wetland eco-hydrological state indicating factors have been divided in three parts like pressure, 161 state and response. to agricultural land map has been derived using the Euclidean distance method considering the 226 fact that an area located at very proximity to agricultural land is more susceptible to degradation 227 or conversion than the area away from agricultural land.  year's images for both periods have been taken for developing built-up area maps. Proximity 237 to built-up area maps have been done in the same way through which proximity to APF was 238 prepared using Eq. 4. NDBI value ranges between -1.0 to 1.0. The threshold value of the built-239 up area is 0.14 and -0.038 for Landsat TM and Landsat OLI respectively instead of 0 as 240 delimited by Pal and Ziaul (2017   inland waters, and the same equation (14) has been used here for computing turbidity levels.  The Secchi disk depth (SDD) is also known as 'water clarity' or 'transparency' in aquatic 356 science, is an important parameter of water quality (Wernand, 2010). Therefore, the SDD has 357 been a routine measurement in field surveys of aquatic environments using a Secchi disk since 358 the 1860s (Secchi, 1864). In recent times Remote sensing data is widely used for retrieving the 359 SDD because of its wide application and large aerial coverage (Carlson, 1977). Field specific 360 data of 20 sites have been used to calibrate SDD from image data and the regression model     into subsets accounting for the outputs, and then the same process is recursive for the subsets.

517
In this process, selecting proper attributes to split T, in order to minimize the expected error at 518 a particular node, requires a splitting criterion, and standard deviation reduction is the expected 519 error reduction, calculated by the following equation (21).        596 The potential supply of Ecosystem service (ES) can be defined as the "hypothetical maximum  The expert-based assessment approach has been applied in this study to calculate the ecosystem 611 service potentiality matrix. Expert are approached and asked to provide the potential score of 612 each ecosystem services using a rating scale ranges from 0 (no potential) and 5 (maximum scores. An average score has been calculated based on expert opinion in different ES groups 617 and finally ESP has been derived from average expert-based opinion scores (table 4).

630
The lower reach of the mainstream is dominated by well-defined seasonal riparian wetland.
Wetland extent is found high during monsoon and post-monsoon seasons triggered by high monsoonal rainfall.  pre-dam condition (Table 5). In post-dam phase this areal extent has increased to 39.36%,     (Table 6).

692
ESP is also used for validating the EHS models. This trend is also true when comparison is done between individual ESP categories with EHS zones (Figure 9).   of the ecosystem is also very dynamic. So, the assessment of wetland EHS of wetlands is a 714 very difficult task, due to the impact of controlling factors that vary spatially and temporally.

715
Remote sensing time series data has been used to analyze EHS mapping because of wide data 716 characteristics. Following the PSR framework, synthetic indices for pressure, state and 717 response indicators, were developed which facilitate a rapid and quantitative assessment of 718 eco-hydrological state.

719
In this study, wetland eco-hydrological state modelling has been carried out following PSR damming after situation, degradation of EHS is also partially explained by these factors.

740
Degradation of EHS has direct impacts on ecosystem productivity and services of the wetland.

741
Based on the author's knowledge ML algorithms using field-based parameters have not been 742 yet applied in previous works for modelling EHS in the world. Multi-model ensemble approach 743 has been applied for predicting EHS in order to obtain a better scope for selecting the model 744 with greater precision. ML models provide scope for objective solution of complicated relations among the multiple parameters (Han et al., 2020). For validating the models apart from statistical approaches, the Ecosystem Service Potential (ESP) approach has been adopted 747 considering 31 numbers of ecosystem service parameters. This approach of validating the EHS 748 model is very novel. As the dataset for developing ESP have been collected from field experts, 749 application of this for validating EHS models could be treated as justified. This method has 750 some demerits since it is based on Likert scale and therefore absolute state is not clearly 751 explained. Moreover, objectivity of the result is another issue that could be raised. However, 752 this method is accepted by the scientific community as it accurately provides relative 753 judgement in different places. This work is a good blend of ML modelling in PSR framework.   junior research fellowship to conduct the research work presented in this paper. 778 The data that support the findings of this study are available from the corresponding author, 779 [swadespal2017@gmail.com], upon reasonable request.